End of training

README.md

@@ -1,9 +1,15 @@
 ---
 base_model: Minbyul/llama2-7b-wo-medication_qa-sft
 tags:
+- alignment-handbook
 - trl
 - dpo
 - generated_from_trainer
+- trl
+- dpo
+- generated_from_trainer
+datasets:
+- HuggingFaceH4/ultrafeedback_binarized
 model-index:
 - name: llama2-7b-dpo-full-sft-wo-medication_qa
   results: []
@@ -14,17 +20,17 @@ should probably proofread and complete it, then remove this comment. -->
 
 # llama2-7b-dpo-full-sft-wo-medication_qa
 
-This model is a fine-tuned version of [Minbyul/llama2-7b-wo-medication_qa-sft](https://huggingface.co/Minbyul/llama2-7b-wo-medication_qa-sft) on an unknown dataset.
+This model is a fine-tuned version of [Minbyul/llama2-7b-wo-medication_qa-sft](https://huggingface.co/Minbyul/llama2-7b-wo-medication_qa-sft) on the HuggingFaceH4/ultrafeedback_binarized dataset.
 It achieves the following results on the evaluation set:
-- Logits/chosen: -0.4292
-- Logits/rejected: -0.4708
-- Logps/chosen: -476.1255
-- Logps/rejected: -635.9033
-- Loss: 0.4682
-- Rewards/accuracies: 0.9250
-- Rewards/chosen: -0.1609
-- Rewards/margins: 0.9415
-- Rewards/rejected: -1.1024
+- Loss: 0.4396
+- Rewards/chosen: -0.1779
+- Rewards/rejected: -1.2468
+- Rewards/accuracies: 0.9500
+- Rewards/margins: 1.0689
+- Logps/rejected: -650.3414
+- Logps/chosen: -477.8221
+- Logits/rejected: -0.4720
+- Logits/chosen: -0.4277
 
 ## Model description
 

all_results.json

@@ -1,5 +1,18 @@
 {
     "epoch": 1.0,
+    "eval_logits/chosen": -0.42766323685646057,
+    "eval_logits/rejected": -0.4719906747341156,
+    "eval_logps/chosen": -477.8221130371094,
+    "eval_logps/rejected": -650.3414306640625,
+    "eval_loss": 0.4395619034767151,
+    "eval_rewards/accuracies": 0.949999988079071,
+    "eval_rewards/chosen": -0.17786642909049988,
+    "eval_rewards/margins": 1.0689051151275635,
+    "eval_rewards/rejected": -1.2467713356018066,
+    "eval_runtime": 37.8151,
+    "eval_samples": 320,
+    "eval_samples_per_second": 8.462,
+    "eval_steps_per_second": 0.264,
     "train_loss": 0.05898992068894947,
     "train_runtime": 471.0672,
     "train_samples": 8410,

config.json

@@ -23,6 +23,6 @@
   "tie_word_embeddings": false,
   "torch_dtype": "bfloat16",
   "transformers_version": "4.39.0.dev0",
-  "use_cache": false,
+  "use_cache": true,
   "vocab_size": 32000
 }

eval_results.json

@@ -0,0 +1,16 @@
+{
+    "epoch": 1.0,
+    "eval_logits/chosen": -0.42766323685646057,
+    "eval_logits/rejected": -0.4719906747341156,
+    "eval_logps/chosen": -477.8221130371094,
+    "eval_logps/rejected": -650.3414306640625,
+    "eval_loss": 0.4395619034767151,
+    "eval_rewards/accuracies": 0.949999988079071,
+    "eval_rewards/chosen": -0.17786642909049988,
+    "eval_rewards/margins": 1.0689051151275635,
+    "eval_rewards/rejected": -1.2467713356018066,
+    "eval_runtime": 37.8151,
+    "eval_samples": 320,
+    "eval_samples_per_second": 8.462,
+    "eval_steps_per_second": 0.264
+}